非负矩阵分解融合高光谱和多光谱数据

由于光谱分辨率和空间分辨率的制约以及物理条件的限制,高光谱数据具有很高的光谱分辨率而其空间分辨率却很低。因此,一般高光谱数据的空间分辨率往往低于仅有几个波段的多光谱数据的空间分辨率。高光谱数据和多光谱数据的融合可以得到同时具有高空间分辨率和高光谱分辨率的数据,进而应用于更高空间分辨率下地物的识别和分类。非负矩阵分解(Nonnegative Matrix Factorization)算法用于实现低空间分辨率高光谱数据和高空间分辨率多光谱数据的融合。首先利用顶点成分分析法VCA(Vertex Component Analysis)分解高光谱数据,得到初始的端元波谱矩阵和端元丰度矩阵;然后用非负矩阵分解算法交替地对高光谱数据和多光谱数据进行分解,得到高光谱分辨率的端元波谱矩阵和高空间分辨率的丰度矩阵;最后两个矩阵相乘得到高空间分辨率和高光谱分辨率的融合结果。在每一步非负矩阵分解过程中,数据之间的传感器观测模型用于分解矩阵的初始化。AVIRIS和HJ-1A数据实验结果分析表明:非负矩阵分解算法有效提高了高光谱数据的所有波长范围内波段数据的空间分辨率,而高精度的融合结果可用于地物的目标识别和分类。

Fusion of Hyperspectral and Multispectral Data Using Nonnegative Matrix Factorization

The hyperspectral data has very high spectral resolution,but low spatial resolution with the trade|off between the spectral and spatial resolution,as well as the physical limits.Therefore,in many cases,the spatial resolution of the hyperspectral image system is lower than the multispectral image system that has the less spectral channels.The fusion of the hyperspectral and multispectral data can produce fused data with high spatial and spectral resolution which can contribute to the identification and classification of the material at a finer spatial resolution.NMF (Nonnegative Matrix Factorization) algorithm is used to perform the fusion of low spatial resolution hyperspectral data and high spatial resolution multispectral data.Firstly,VCA (Vertex Component Analysis) is used to perform the factorization of hyperspectral data to obtain endmember spectra matrix and abundance matrix; Secondly,NMF(Nonnegative Matrix Factorization) is used to unmix the hyperspectral and multispectral data alternatively to obtain high spectral resolution endmember spectra matrix W and high spatial resolution abundance matrix H.Finally,fusion data with high spectral resolution and high spatial resolution can be obtained by multiplying the two matrices,.Sensor observation models of the data are built in the initialization matrix of each NMF unmixing procedure.The experiments with AVIRIS data and HJ-1A data have shown that NMF method can be used to improve the spatial resolution on all wavelength regions,and the higher qualities of the estimated data by NMF can be used for the classification of the materials and identification at a finer spatial resolution.